Openclaw Update: Agents, Media Generation, Providers, Memory & UI

There’s a ton of information about the latest update from Openclaw.
Here’s a brief summary:

🔥 Key highlights
Deprecated configuration fields have been cleaned up → migration to new standard paths.
Video and music generation added directly via agents.
Provider integrations improved (Qwen, Fireworks, Bedrock, etc.).
Multilingual UI added (including Ukrainian).
(https://docs.openclaw.ai/uk) — this is just beautiful; not everything is translated yet, but there’s already a big foundation

🤖 Agents and tools
New video_generate and music_generate.
(https://docs.openclaw.ai/uk/tools/video-generation) — this is also a banger
ComfyUI support for local media generation.
Claude CLI integrated via MCP (better streaming and stability).
Improved task execution progress (structured steps).

⚙️ Providers
Bedrock Mantle added + auto-config via AWS.
New models and integrations (MiniMax, Ollama Search, TTS).
OpenAI: support for the new GPT-5 models.

🧠 Memory and search
Introduced a “dreaming” system (light/deep/REM phases).
Better long-term knowledge retention.
New tools for memory analysis.
Control over data “aging.”

⚡ Cache and performance
Improved prompt caching → lower costs and faster responses.
Stable fingerprints for cache reuse.
Cache diagnostics via openclaw status.

🧩 UI and plugins
ClawHub integrated into the UI (search and install).
Better onboarding via TUI.
Plugins can be reinstalled without dangerous flags.

🔐 Security and control
Context visibility settings (allowlist).
Improved exec approvals (iOS + Matrix).
Request overrides (headers, proxy, TLS).

🧹 Other
Claude CLI backend removed (deprecated).
Config and documentation simplified.
Improved runtime and workflow execution stability.

If you also want to keep track, you can go here:
https://github.com/openclaw/openclaw/releases